RA 05h 35m 17.3s | Dec -05° 23′ 28″ | L2 Halo Orbit
| UTC Timestamp | Source Coordinates | Receiver Config | Frequency Peaks | Data Archive | Analysis Notes |
|---|---|---|---|---|---|
| 2025-01-15T03:47:00Z |
Llama-3.1-8B-Instruct
Meta AI · Open Weights
|
RTX 3090 — 24GB
Ollama v0.4.0
Q4_K_M compression |
MMLU-Pro
GPQA
IFEval
|
Download FITS | Strong detection. 14.2GB peak utilization. 47 tok/s sustained throughput. Six-hour continuous observation. Quantization noise floor acceptable. SNR: +24 dB |
| 2025-01-12T22:15:00Z |
Qwen2.5-14B-Instruct
Alibaba DAMO Academy
|
RTX 3090 + 64GB host
llama.cpp b4000
Q5_K_M · CPU offload |
HumanEval
MBPP
MultiPL-E
LiveCodeBench
|
Download FITS | Moderate detection. Eight layers CPU-offloaded. 23 tok/s with interference. Code-generation band exceptionally clear. Rust frequency truncated by window limits. SNR: +12 dB |
| 2025-01-08T17:33:00Z |
Mistral-Small-24B-Instruct
Mistral AI · Paris
|
Dual RTX 3090 NVLink
vLLM 0.6.5
BF16 native · no compression |
MATH
GSM8K
BBH
|
Download FITS | Extraordinary detection. Native 24B uncompressed. 89 tok/s — highest recorded. Chain-of-thought coherence exceeds all compressed sources. NVLink crosstalk negligible. SNR: +31 dB |
Extended observations of roleplaying phenomena, creative emission spectra, character coherence mapping, and experimental signal processing at sensitivity levels below standard detection thresholds.
Repoint DishSETI-Alpha · Neocities Relay Station · No signal modulation detected from host
Continuing observation · EST. 2025